ONNX
Interchange format for models, so a trained model runs on another runtime.
Overview
ONNX is a file format and operator set that a trained model can be exported to and another runtime can execute. It solves the problem that a model is inseparable from the framework that produced it, so serving it means shipping that framework. It exists because separating the model from the training code is what allows a model to be optimised and run by runtimes that never saw its original definition.
Technology Graph
How ONNX connects to the rest of the stack.
Applied to
- Inference concept Infrastructure ONNX used for Inference High ✓ source
Tools
Software that implements or integrates with ONNX.